Research On Target Search And Location Of Mobile Robot Based On Monocular Vision

With the rapid development of robotics and artificial intelligence,visual positioning has become a hot topic in the field of robotics research.The vision sensor can provide information such as distance and orientation for the robot.It is an important source for the robot to obtain information about the surrounding environment.The monocular vision sensor is simpler and easier to calibrate than a binocular vision sensor.In this dissertation,the target positioning of the mobile robot's monocular vision system is studied in combination with the actual scene.Finally,the measured position information is used to control the robot to move to the target.On the basis of reviewing the current situation of mobile robot positioning algorithm and visual positioning research,the main work done in this dissertation completed is as follows:(1)Camera calibration.This topic uses the MATLAB calibration box to perform parameter calibration on the CanonVC-C50i camera.This method is simple and clear,and has high precision.However,the accuracy of the calibration box has to meet two major characteristics:it requires a large number of calibration pictures and accurate corner extraction.This dissertation takes respectively 4 and 20 calibration images for testing,And particle swarm optimization algorithm is used to optimize the internal parameters obtained from the four images.Experimental results show effectiveness of the method.(2)Target position.From the versatility of the algorithm,put forward two target location algorithms based on the circular marker.The results show that the second algorithm has the characteristics of high precision and fast detection in the complex background.But the second algorithm still has some limitations.If there is a circular interference with the same color as the marker,then a test with other colors marker is required.This also makes the method attractive in applications such as target location,grasping,and target tracking.(3)Monocular vision ranging.In this dissertation,a distance measurement model based on spatial geometric plane constraints is used.The model measures the absolute distance between the robot camera and the target without experience.The results of two sets of ranging experiments show that the algorithm is feasible,the error of short distance measurement is controlled within 1cm,and the distance error of long distance measurement can be controlled within 3cm.(4)Robot target search and positioning experiment.Target search and location experiments are based on images.The subject makes the relevant target search strategy based on the actual experimental environment.And Pioneer3-DX robot platform and analysis of the robot program execution process are researched.According to the position information obtained from the distance measurement,adjusting the robot to move to in the front of the target.The accuracy of the algorithm is verified by the difference between the position of the robot after the movement and the target position.The experimental results show that the target positioning effect of mobile robot in this dissertation is very good,which greatly improves the positioning accuracy and the success rate is high.And robot action is more intelligent,without human intervention can be automatically set to complete the task.